Dear Splus users,
I am trying to run GLMs on fisheries data, which often is best described by
a Aitchison's Delta distribution - that is, there is a substantial
probability of zero values (ie zero fish catch), and the non-zero values are
log-normally distributed. These data are of course awkward: eliminating
zeros enables Gamma(link=log) functions to fit, but this is a biased part of
the dataset. In some instances binomial error models are ok, but never give
a good fit.
Does anyone know how to deal with delta distributions in GLM/GAM models?
Thanks,
David Agnew
------------------------------------------------------------------------
Dr David Agnew | |
Research Fellow | |
Renewable Resources Assessment Group | o /(
T. H. Huxley School of Environment, | o {[[[[\ /
Earth Sciences and Engineering | >{o[[[[[[}<|(
Imperial College | | {[[[[/ \
8 Prince's Gardens | d.agnew@ic.ac.uk | \(
London SW7 1NA |tel:+44(0)207 5949273 |
UK |fax:+44(0)207 5895319 |
------------------------------------------------------------------------
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
|